一种基于地址-事件表达的实时视觉传感器实现方法

为获取高帧频、大动态范围、低数据量的精确视觉信息,提出了一种基于地址-事件表达(AER)的实时视觉CMOS传感器实现方法。采用AER方式,多模式的行仲裁及实时时间标记,有效减小读出数据量,减小时域行间信息扭曲;利用像素级光强变化感知电路探测光强变化,双采样脉冲宽度调制(PWM)电路量化光强。仿真结果显示,在100lx光强下,获得最小等效帧频1000frame/s,10lx光强下为100frame/s,静态动态范围大于133dB,视频动态范围48.16dB,并可实现实时时间标记、多模式仲裁的功能,减小输出数据量11.61%～42.74%。结果表明,该方法可完成实时视觉信息的采集、处理、输出,适用于高速、高动态范围视觉领域。

An Implementation Method of Real-Time Vision Sensor Based on Address Event Representation

In order to achieve high frame frequency, high dynamic range (DR) and low data quantity of visual information accurately, an implementation method of real-time vision sensor based on address-event representation is proposed. The readout data quantity and time-based information distortions decrease, benefitting from AER, adjustable row arbitration and time stamp. The method can detect light-intensity change by changing detector circuits and quantify the light intensity changes by double sampling pulse width modulation (PWM) circuits. The simulated results show that in the illumination conditions of 100 lx and 10 lx, the minimum equivalent frame frequency is 1000 frame/s and 100 frame/s, respectively. DR can reach 133 dB, and vision DR is 48.16 dB. The output can be decreased by 11.61%~42.74% compared with previous one. It proves that the proposed method can perform real-time optical signal capturing, processing and readout as well as be applied to the field of high-speed and high DR image and vision.